Integrative Computational Approaches to Drug Repurposing for Alzheimer’s Disease: Leveraging Multi-Omics, Electronic Health Records, and Generative AI

Abstract

Alzheimer's disease (AD) involves complex, multifactorial brain changes that begin decades before symptom onset, making the development of therapeutic and preventive interventions extremely challenging. With AD prevalence increasing worldwide, the development of safe and effective therapies has become an urgent public health priority. Drug repurposing, the identification of new therapeutic uses for existing drugs, represents a promising strategy for accelerating drug development in AD, offering advantages through established drug safety profiles, lower costs, and reduced development timelines. Although hundreds of AD repurposing candidates have been proposed over the past decade, few have undergone rigorous validation, making it difficult to prioritize candidates for clinical investigation. To bridge this gap, we applied three approaches integrating diverse sources of -omics and clinical data to suggest promising drug repurposing candidates for AD: (1) leveraging large language models to rapidly mine and synthesize the biomedical literature, (2) performing transcriptomic analysis to identify drugs capable of reversing AD-associated changes in gene expression, and (3) using Mendelian randomization to identify drugs acting on proteins causally associated with AD. We then investigated the real-world effects of the candidate drugs using data from electronic health records and national health insurance claims. Our findings support aspirin, metformin, losartan, and simvastatin as high-priority candidates warranting further evaluation in randomized clinical trials for AD. This work not only advances our understanding of AD repurposing opportunities, but also presents a flexible, generalizable computational framework applicable to a broad range of complex diseases.